Sensitivity of tropical cyclone Jal simulations to physics parameterizations R Chandrasekar and C Balaji∗ Department of Mechanical Engineering, Indian Institute of Technology, Madras, Chennai 600 036, India. ∗Corresponding author. e-mail: [email protected] In this study, the sensitivity of numerical simulations of tropical cyclones to physics parameterizations is carried out with a view to determine the best set of physics options for prediction of cyclones originating in the north Indian Ocean. For this purpose, the tropical cyclone Jal has been simulated by the advanced (or state of science) mesoscale Weather Research and Forecasting (WRF) model on a desktop mini super computer CRAY CX1 with the available physics parameterizations. The model domain consists of one coarse and two nested domains. The resolution of the coarse domain is 90 km while the two nested domains have resolutions of 30 and 10 km, respectively. The results from the inner most domain have been considered for analyzing and comparing the results. Model simulation fields are compared with corresponding analysis or observation data. The track and intensity of simulated cyclone are compared with best track estimates provided by the Joint Typhoon Warning Centre (JTWC) data. Two sets of experiments are conducted to determine the best combination of physics schemes for track and intensity and it is seen that the best set of physics combination for track is not suitable for intensity prediction and the best combination for track prediction overpredicts the intensity of the cyclone. The sensitivity of the results to orography and level of nesting has also been studied. Simulations were also done for the cyclone Aila with (i) best set of physics and (ii) randomly selected physics schemes. The results of the Aila case show that the best set of physics schemes has more prediction skill than the randomly selected schemes in the case of track prediction. The cumulus (CPS), planetary boundary layer (PBL) and microphysics (MP) parameterization schemes have more impact on the track and intensity prediction skill than the other parameterizations employed in the mesoscale model. 1. Introduction mesoscale weather events. The prediction of TCs using mesoscale NWP models is highly influ- Tropical cyclones (TC) are important weather phe- enced by initial/boundary conditions and physics nomena that cause heavy wind, torrential rain parameterizations employed in the model. In view and enormous damage to life and property when of this, ‘fixing’ the one best combination of they cross the coastal regions. Accurate pre- physics schemes in the mesoscale NWP mod- dictions of TC track and intensity with timely els for the whole globe is formidable. Further, warning to people will minimise loss of life and the NWP model performance is very sensitive to property. Availability of state-of-the-art numeri- grid sizes and the geographical region of inter- cal weather prediction (NWP) models along with est. The best set of schemes for one region may efficient computing facilities to carry out simu- not be suitable for some other region. The horizontal lations has made possible the prediction of the and vertical resolutions are also very important Keywords. Cyclones; track and intensity prediction; WRF; physics parameterizations. J. Earth Syst. Sci. 121, No. 4, August 2012, pp. 923–946 c Indian Academy of Sciences 923 924 R Chandrasekar and C Balaji factors, since different parameterization schemes scheme for CPS, Pleim-Xu scheme for land surface can give different results for different model reso- and asymmetrical convective model 2 (ACM2) lutions. Sensitivity experiments are the only log- scheme for PBL produces better surface wind pre- ical way to identifying the best set of physics diction agreement with both station observations schemes for a particular region. Srinivas et al. and Quick Scatterometer(QuikSCAT) surface wind (2007) conducted a sensitivity study of the Andhra data. The above studies on regional weather mod- severe cyclone (2003) by using the National Cen- els clearly indicate that the best combinations of ter for Atmospheric Research (NCAR) fifth gen- physics parameterization schemes are very essen- eration non-hydrostatic mesoscale model (MM5). tial to predict a mesoscale weather system. Previ- They reported that the planetary boundary Layer ous sensitivity studies on mesoscale models were (PBL) schemes and convective parameterization mostly initiated with NCEP Final analysis (FNL) schemes (CPS) play an important role in predict- data as initial and boundary conditions. Very few ing both the intensity and movement of the model studies have been initiated with the NCEP Global simulated storms. They also concluded that the Forecast System (GFS) real time prediction. In combination of Mellor Yamada (MY) scheme for this study, the sensitivity of a very recent tropical PBL parameterization and Kain-Fritch 2 (KF2) cyclone Jal to model physics parameterization is scheme as the CPS give the best results in terms conducted with GFS real time predictions as the of intensity and track. A sensitivity study of the initial and boundary conditions. Use of GFS real Orissa super cyclone by Rao and Prasad (2007) time prediction is akin to running a real time fore- also indicates that a combination of the MY scheme cast. Furthermore, most of the previous sensitiv- from the PBL and KF2 from CPS gives better ity studies have been done with only combinations results in terms of cyclone track and intensity pre- of Cumulus, PBL and microphysics parameteriza- diction. Deshpande et al. (2010) however conclude tions and very few studies consider different radi- that the prediction of cyclone track and inten- ation schemes and other physics options. In this sity are highly sensitive to only convective param- study, all physics parameterizations available in the eterization schemes compared to other physical ARW v 3.1.1 model are systematically evaluated. parameterization schemes based on their studies on cyclone Gonu using the MM5 model. Loh et al. (2010) also report that the PBL parameterization 2. Synoptic history of Jal schemes do not significantly affect track and inten- sity prediction of near equatorial typhoons. Prater Cyclone Jal was the fourth severe cyclonic storm and Evans (2002) modeled the tropical cyclone Irene of the north Indian Ocean in 2010. Jal developed (1999) with various CPS and concluded that the from a low pressure area in the South China Sea Kain-Fritsch (KF) scheme in MM5 produces rel- and organized into a tropical depression on October atively accurate storm predictions compared to 28. Late on November 1, the tropical depres- observations. Mandal et al. (2004) simulated two sion crossed the Malay peninsula, and entered severe tropical cyclones in the Bay of Bengal the extreme eastern part of the Indian Ocean, using MM5 with various CPS, PBL and radia- and moved continuously towards the west and tion schemes and inferred that the combination of entered the Bay of Bengal on 4 November 2010. Medium Range Forecast (MRF) scheme in PBL The system further strengthened on November 4 and Grell scheme in CPS with Community Cli- 18 UTC (Coordinated Universal Time) and the mate Model Version 2 (CCM2) scheme in radia- JTWC declared it as a tropical storm with a max- tion gives better performance in so far as track imum wind speed of 19.16 m/s. Later the storm and intensity predictions are concerned. Yang and moved west, gained energy from the warm waters Ching (2005) also concluded that the MRF in PBL and strengthened again as a category 1 storm on and Grell in CPS combined with the Goddard November 6 00 UTC and continued till Novem- Graupel in cloud microphysics scheme give the best ber 7 06 UTC as a category 1 storm. The system performance in the study of typhoon Toraji (2001). weakened and became a deep depression after its Based on a study of impact of cloud microphysics landfall near Chennai at 18 UTC on November 7. on hurricane Charley, Pattnaik and Krishnamurti (2007) reported that the microphysical para- meterization schemes have strong impact on the intensity prediction of hurricane but have negligi- 3. Description of the model ble impact on the track forecast. Krieger (2009) analyzed the sensitivity of Advanced Research The model used in this study is the Advanced WRF (ARW) v 3.0 model on the Beaufort Sea Research WRF (ARW) v 3.1.1 model, a widely region with the available physics parameterization used community mesoscale model developed by schemes and concluded that the combination of KF NCAR. The ARW is a next generation of NCAR Sensitivity of Jal simulations to physics parameterizations 925 Table 1. Model physics parameterization schemes. fifth generation mesoscale model (MM5) which incorporates recent and new physics parameteriza- Cumulus tion schemes including some of the schemes from 1. Kain-Fritsch(new Eta) scheme KF the MM5 model. The available physics parameter- 2. Betts-Miller-Janjic scheme BMJ ization schemes in the ARW model are detailed 3. Grell-Devenyi ensemble scheme GD in NCAR technical notes by Skamarock (2008). 4. New Grell scheme GRE The ARW can be used for both research and PBL operational applications. It is a non-hydrostatic 1. Yonsei University scheme YSM mesoscale model with run-time hydrostatic option. 2. Mellor-Yamada-janjic(Eta) TKE scheme MYJ The model dynamic solver integrates the flux 3. Quasi-Normal Scale Elimination QNSE form of compressible non-hydrostatic Euler equa- (turbulence theory) tion. The variables in this equation are conserva- 4. Mellor-Yamada-Nakanishi-Niino MYNN2.5 tive properties following the philosophy of Ooyama 2.5 level TKE (1990). The equations are formulated using a ter- 5. Mellor-Yamada-Nakanishi-Niino MYNN3 rain following mass vertical co-ordinate following 3 levelTKE Laprise (1992). In the ARW solver, a time-split 6. Asymmetrical Convective Model ACM2 integration scheme is used for solving the com- version 2 pressible non-hydrostatic Euler equation.